Anonymous Vehicle Detection for Secure Campuses: A Framework for License Plate Recognition using Deep Learning

Abstract

Automatic license plate recognition is being widely used for numerous applications since its inception. The ability to procure license plate numbers accurately has been beneficial in maintaining traffic rules, parking enforcement, and security. In this paper, we have discussed the results of using ALPR for recognition of anonymous vehicles entering our university campus. We used deep learning for license plate localization and Tesseract OCR for license plate recognition. By doing so we could read the license plates of vehicles entering a particular campus and verify if the vehicle is authorized by comparing it with a predefined list of authorized vehicles. To efficiently extract these number plates we have trained our model using Faster RCNN and tuned it to get the best output. The results of which have been discussed in this paper. Further, the image processing techniques used for preprocessing the identified number plate have been mentioned here. For character segmentation and character recognition, we have used tesseract. While training our model for number plate extraction the minimum loss obtained was 0.011 with RMSprop optimizer at initial learning rate 0.002.